A bed making quality detection method for a hotel bed making service robot

Abstract

This invention discloses a method for inspecting the quality of bed making in hotels using a bed-making service robot. The method is integrated into the robot's control system and includes the following steps: determining the bed-making quality inspection target, defining image range information, range type, and inspection dimension; generating control commands for the robot's sensors based on the image range information; acquiring image group information returned by the robot sensors; loading an image recognition model, including a standard definition unit, a feature extraction unit, and a feature comparison unit; inputting the image group information into the image recognition model to obtain inspection results and correction strategies. According to this technical solution, when inspecting the quality of bed making services provided by the robot, clear inspection standards can be established, reducing reliance on manual labor, saving labor costs, improving the efficiency of hotel room preparation and control, ensuring greater accuracy and consistency in inspection, and providing readable evaluation results.

Description

Technical Field

[0001] This invention relates to the field of robotics, and more specifically, to a method for inspecting the quality of bed making for a hotel bed-making service robot. Background Technology

[0002] In hotel services, bed-making quality inspection is a crucial step in ensuring room service quality. This involves judging whether sheets and duvet covers are flat, measuring sheet centerline deviation and pillow spacing to ensure they meet standards (e.g., Western-style bed-making requires a centerline deviation of ≤1cm), and checking for stains, damage, or hair on the bedding surface. With advancements in artificial intelligence and robotics, the hotel industry is gradually introducing robots to replace manual labor in repetitive tasks like bed-making, and quality inspection methods are shifting from manual to automated checks. However, existing automated inspection technologies often rely on single sensors (such as infrared rangefinders for bed height differences), resulting in low efficiency, insufficient accuracy, and susceptibility to environmental interference, making it difficult to achieve multi-dimensional, high-precision comprehensive evaluation.

[0003] For example, patent CN111680915A discloses a hotel room management method that uses a service robot to take photos of the room to check if the furnishings are up to standard, but it does not involve automatic detection of bed-making quality, nor is it integrated with the control system of the bed-making robot. There is an urgent need for an efficient, accurate quality detection method for hotel bed-making service robots that can be seamlessly integrated with the control system to improve the level of service intelligence. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method for inspecting the quality of bed making in a hotel bed-making service robot. This method is integrated into the control system of the hotel bed-making service robot and includes the following steps:

[0005] The quality inspection targets for bed making are determined, and the image range information, range type, and inspection dimensions are determined based on these targets. Range types include: bed side, sheet folds, bed surface, and pillows. Inspection dimensions include: neatness, flatness, and cleanliness. The image range information is determined based on the range type and represents the operational characteristics of the robot's sensors when acquiring environmental information. The image range information includes range information identifier, sensor position, focus position, sensor image angle, and range type.

[0006] Generate control commands for the robot's sensors based on the image range information;

[0007] Acquire image group information returned by the robot's sensors; image group information includes: image group information identifier, command identifier, image sequence number, and image file;

[0008] Load the image recognition model; the image recognition model includes a standard definition unit, a feature extraction unit, and a feature comparison unit; the standard definition unit is used to extract standard attributes from region rules, the standard attributes include the three-dimensional position of the bed's contour edge marker points, feature point recognition rules, and quality assessment rules; the feature extraction unit is used to identify feature points from the specified image; the feature comparison unit is used to calculate the detection results based on the feature points, and supports redefining the image range information during the calculation process;

[0009] Input image group information into the image recognition model to obtain detection results and correction strategies.

[0010] The control system of the hotel bed-making service robot controls the behavior of the robot's components through a behavior execution module, a sensor module, and a control module.

[0011] The components include a mobile base, a robotic arm, and an end effector;

[0012] The behavior execution module includes an end effector unit, a robotic arm control unit, and a motion control unit, which are deployed on the constituent components to respond to control commands and control the actions of each constituent component.

[0013] The sensor module includes a vision system unit consisting of an RGB camera and a structured light camera, which is placed on the robot's components and can collect environmental information at a specified location, angle, and pattern.

[0014] The control module includes an operation specification library, a task planning engine, and a mode switching unit. It is deployed on the control server and is used to plan the specific operation steps for bed making based on environmental information, generate functional instructions, and send control instructions to the behavior execution module.

[0015] The control command includes: command identifier, execution module ID, operation component ID, target point location, operation type, and control status; when the operation type is shooting, the control status includes camera position, camera orientation, focus position, and shooting parameters.

[0016] Furthermore, when determining the image range information and range type, the target for bed making quality inspection is decomposed into positions, and a focal position corresponding to the range type is generated according to the region rules; the focal position is determined by three-dimensional coordinates based on the mattress coordinate system; one focal position corresponds to multiple sensor shooting angles.

[0017] Before generating control commands for the robot's sensors, the process of generating control commands is optimized. The optimization goal is to reduce repetitive operations of the robot. The optimization method is to merge the contents of multiple image range information and store the correspondence between control commands and image range information.

[0018] During optimization, it is calculated whether there is an overlap between the focal positions corresponding to all image range information. If there is an overlap and the sensor shooting angle is the same, the control commands of multiple image range information are merged into one control command. If there is no overlap, under the premise of ensuring the clarity of each focal position, the degree of concentration of the focal positions is judged, the core concentration point of multiple focal positions with close distance is found, the core concentration point is taken as the optimal focal position, and the sensor shooting angle is added to generate the control command. Other focal positions with close distance do not need to generate control commands.

[0019] Furthermore, after acquiring the image group information returned by the robot's sensors, the complete image group information that perfectly corresponds to the image range information is restored based on the correspondence between the control commands and the image range information.

[0020] The feature extraction unit identifies feature points from a specified image by including the following steps:

[0021] Perform image optimization processing on the specified image to generate an image to be evaluated;

[0022] Feature points are identified based on detection dimensions and standard attributes, including neatness feature points, flatness feature points, and cleanliness feature points;

[0023] Among them, the neatness feature points are used to determine whether the pillow and bed sides are aligned and whether the folds of the bed sheet meet the standards; the flatness feature points are used to determine whether the wrinkled areas of the bed surface meet the standards; and the cleanliness feature points are used to determine whether there is damage or foreign objects in the area corresponding to the range type.

[0024] When the feature comparison unit calculates the detection result based on feature points, it includes the following steps:

[0025] Select the detection dimension, and extract feature points and their 3D positions from the image to be evaluated based on the detection dimension;

[0026] Extract the standard attributes of quality assessment rules;

[0027] Quality inspection is performed based on standard attributes, feature points, and detection dimensions. When the detection dimension is flatness, if a suspected wrinkled area is detected, the area is used as a flatness feature point as the focus position to redefine the image range information. When the detection dimension is cleanliness, if a suspected foreign object is detected, the area is generated as a cleanliness feature point as the focus position to redefine the image range information.

[0028] Output the detection results, which include the detection dimension, focus position, and quality status.

[0029] Furthermore, if the detection dimension is neatness, the standard attributes also include the offset standard value of the bed-making quality detection target; if the detection dimension is flatness, the standard attributes also include the fold density value and fold size value of the bed-making quality detection target; if the detection dimension is cleanliness, the standard attributes also include the size and quantity of foreign objects in the bed-making quality detection target.

[0030] Among them, the correction strategy refers to further generating control instructions for the robot to confirm the detection results and the quality of the correction.

[0031] When confirming the detection results, the correction strategy is to redetermine the image range information and then generate action sequences and control instructions based on the image range information; when correcting the quality, the correction strategy is to generate action sequences and control instructions to control the robot's behavior, and control the robot to perform actions such as smoothing out wrinkles or separating contaminated linens.

[0032] Furthermore, after the correction strategy is executed a specified number of times, the quality inspection ends and an inspection report is generated. The inspection report indicates the type, quantity, and specific location of the problematic bedding, and clarifies the key points that need to be reworked and rectified.

[0033] This invention reduces reliance on manual labor, saves labor costs, and improves the efficiency of hotel room preparation. In terms of quality control, it ensures greater accuracy and consistency in inspections, improving the accuracy of stain identification and reducing measurement errors in location deviation. Simultaneously, it provides readable assessment results, offering quality inspection reports that intuitively present the type, quantity, and location of problematic bedding, facilitating hotel room service management. Furthermore, the control process during quality inspection has been optimized, improving control efficiency. This invention embeds bedding quality inspection data into all aspects of hotel operations, enabling cross-departmental data sharing and business collaboration, using digital means to drive improvements in hotel service quality and operational efficiency. In conclusion, the bedding quality inspection system provided by this invention effectively enhances the intelligence level of hotel bedding service robots, offering new ideas for the hotel industry in terms of intelligent upgrading and refined management. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating the bed-laying quality inspection method according to an embodiment of the present invention. Detailed Implementation

[0035] This invention provides a method for inspecting the quality of bed making using a hotel bed-making service robot. This method is integrated into the control system of the hotel bed-making service robot, sharing a control platform with the robot. It utilizes the robot's existing sensors to acquire environmental information and automatically performs inspections during or after the robot has laid out the bedding. To address the aforementioned problems, this invention provides a method for inspecting the quality of bed making using a hotel bed-making service robot. In the application environment of this invention, the control system controls the behavior of the robot's components through a behavior execution module, a sensor module, and a control module. The robot's components include at least a mobile base, a robotic arm, and an end effector. The behavior execution module, comprising an end effector unit, a robotic arm control unit, and a motion control unit, is deployed on the robot's components to respond to control commands and control the actions of each component. The sensor module includes a vision system unit composed of devices such as RGB cameras and structured light cameras, placed on the robot's components, capable of acquiring environmental information at specified locations, angles, and modes. As the core of the control system, the control module includes an operation specification library, a task planning engine, and a mode switching unit, deployed on a control server, used to plan specific operation steps for bed-laying based on environmental information, generate control commands, and send control commands to the behavior execution module.

[0036] The control command specifically includes the command identifier, execution module ID, operation component ID, target point location, operation type, and control status. Among them, the operation component ID is the component identifier of each component controlled by the behavior execution module, the target point location is the coordinate value in the dynamic map of the room, the operation type includes actions that the component can respond to, such as grabbing, relaxing, moving, vibrating, and shooting, and the control status is the specific parameters of the operation type. For example, when the operation type is shooting, the control status includes the camera position, camera orientation, focus position, shooting parameters, etc. For the coordinated operation of multiple components, that is, when multiple components need to execute at a specified time difference, the control status also includes time parameters.

[0037] In the control module of this invention, the detection function is implemented using Python (OpenCV / PyTorch). Quality judgment rules are pre-stored in the operation specification library. Action sequences are generated through the task planning engine. The mode switching unit generates control instructions and sends them to the behavior execution module of the bed-laying robot. After completing the detection based on the returned environmental information, the control instructions for correcting the quality are sent to the nodes, forming a complete closed loop of "bed-laying-detection-correction".

[0038] The bed-making quality inspection method provided by this invention is as follows: Figure 1 As shown, it includes the following steps:

[0039] Step S100: Determine the bed-making quality inspection target. Based on the bed-making quality inspection target, determine the image range information, range type, and inspection dimensions. The bed-making quality inspection target includes the bed and items on the bed (such as pillows). According to the hotel's bed-making service standards, the range types are defined as bed sides, sheet folds, bed surface, and pillows. The inspection dimensions include neatness, flatness, and cleanliness. The inspection dimensions are determined by the environmental information corresponding to the range type. The neatness of the sheet is determined by collecting environmental information (images) for the range types of bed sides and sheet folds. The placement of the pillows relative to the bed sides (such as the headboard) in the images is used to determine whether the placement is compliant. The flatness of the bed surface is determined by identifying wrinkles in the environmental information (images). Foreign objects and stains are identified in the images corresponding to all range types to determine the cleanliness.

[0040] Image range information is determined based on the range type and represents the operational characteristics of the control sensor when acquiring environmental information. When the control module generates control commands, it converts the image range information into the control state of the commands (such as camera position, camera orientation, focus position, and shooting parameters). After issuing the control commands to the robot, the sensor module can acquire images related to environmental information at a specified position and in a specified manner according to the requirements of the control state. Specifically, image range information includes range information identifiers, sensor position, focus position, sensor shooting angle, range type, etc. The control system generates one or more control commands based on the image range information, controlling the robot to extend the end effector carrying the specified sensor (such as an RGB camera and a structured light camera) to the specified position and acquire the image file according to the control state.

[0041] The process of making the bed involves many steps that require inspection. Even judging the quality of a single step requires images from multiple positions and angles. Different types of quality inspections first require determining the inspection location. This is done by pre-determining the optimal shooting points using a large number of images. The data related to the shooting points and methods are then pre-stored as area rules in the control system's operation specification library. These area rules also include standard attributes, such as the 3D position of the bed's outline markers, feature point recognition rules, and quality assessment rules.

[0042] Furthermore, it is necessary to decompose the target location for bed-making quality inspection and generate a focal position corresponding to the range type based on the optimal shooting point pre-set in the area rules; the focal position is determined by three-dimensional coordinates based on the mattress coordinate system. In practical applications, inspecting a step requires images from multiple points and angles to evaluate quality. Therefore, one range type corresponds to multiple focal positions, and one focal position also corresponds to multiple sensor shooting angles.

[0043] For example, when inspecting the smoothness of bed sheet edges, the range type is bed side, and the content defined in the area rules includes:

[0044] Shooting point 1: 45 degrees to the outside of the foot of the bed, at the same height as the bed surface, 30cm away from the edge of the bed;

[0045] Shooting point 2: 2 / 3 of the way down from the foot of the bed, directly above, 50cm from the bed surface;

[0046] Extracted features: edge segment curvature, length, and tilt angle;

[0047] Judgment rules: radius of curvature > 1m, length deviation < 2cm, tilt angle < 5°;

[0048] When detecting the symmetry of pillow placement, the range type is pillow, and the area rule can be defined as:

[0049] Shooting location: Directly in front of the head of the bed, at the same height as the pillow, 50cm away from the head of the bed;

[0050] Feature extraction: Center coordinates and angle of the pillow rectangle;

[0051] Judgment rules: The angle between the center line and the edge of the bed is 90°±2°, and the distance between the two centers from the center line of the bed is ±5cm.

[0052] The way regional rules are represented can be agreed upon through an agreement, such as:

[0053] {

[0054] "Inspection Item": "Alignment of the center line of the bed sheet"

[0055] "Detection Target": "Bed"

[0056] "Scope type":"bed surface",

[0057] "Detection Dimension": "Straightness"

[0058] "Shooting point coordinates":(x,y,z), / / Based on mattress coordinate system

[0059] "Camera Orientation":(α,β,γ), / / Euler angles define pitch / yaw / rotation

[0060] "Shooting parameters":{

[0061] Resolution: 3840×2160

[0062] "Lighting compensation":"Automatic",

[0063] Depth of field range: 0.5-1.2m

[0064] },

[0065] "Feature point definition":[

[0066] {"Name":"Starting point of the centerline","Coordinate range":(x1±Δx,y1±Δy,z1±Δz)},

[0067] {"Name":"End point of the centerline","Coordinate range":(x2±Δx,y2±Δy,z2±Δz)} ]

[0069] }

[0070] Step S110: Generate control instructions for the robot's sensors based on the image range information; in the control module of the robot's control system, a robot action sequence is constructed based on the image range information, and then control instructions are generated to control the robot to take pictures in a specified manner. The robot moves to the specified position according to the control instructions, controls the robotic arm to drive the camera to take pictures at the specified angle and parameters, generates image group information, and sends it back to the control system. However, the content of the image files corresponding to multiple image range information may be repeated, which will cause the robot to perform repetitive operations.

[0071] Therefore, before the control module generates control commands, the process of generating control commands is optimized: the optimization goal is to reduce repetitive operations of the robot, and the optimization method is to merge the contents of multiple image range information and store the correspondence between control commands and image range information according to the command identifier and the image range information identifier.

[0072] During optimization, the system calculates whether there is overlap between the focal positions corresponding to each image range. If overlap exists and the sensor's shooting angle is the same, the control commands for merging the image range information are combined into a single control command, and the relationship between this command identifier and the image range information is stored. If no overlap exists, while ensuring the sharpness of each focal position, the system determines the concentration of the focal positions and then accurately identifies the core concentration points of multiple closely spaced focal positions from all focal positions. These core concentration points are designated as the optimal focal position, and control commands are generated by incorporating the sensor's shooting angle condition. Control commands are not generated for other closely spaced focal positions, thus improving control efficiency. After optimization, the system stores the relationship between this command identifier and the image range information corresponding to the optimal focal position and closely spaced focal positions.

[0073] Step S120: Obtain the image group information returned by the robot sensor; the image group information includes: image group information identifier, command identifier, image sequence number, and image file; since the control command was optimized in step S110, after obtaining the environmental information returned by the robot sensor, the complete image group information that completely corresponds to the image range information is restored according to the relationship between the command identifier and the image range information. Therefore, in the restored complete image group information, one focal position and one shooting angle correspond to multiple image files, and one image group information identifier corresponds to multiple range information identifiers, and one range information identifier also corresponds to multiple image files, and there are cases where the image files are the same.

[0074] Step S130: Load the image recognition model;

[0075] Image recognition models include standard definition units, feature extraction units, and feature comparison units;

[0076] 1) Standard definition units are used to extract standard attributes from the operation specification library of the control module; for example, the feature point recognition rules for detecting the flatness of the edge of the bed sheet are: the curvature, length, and tilt angle of the edge line segment, and the quality assessment rules are: the radius of curvature is >1m, the length deviation is <2cm, and the tilt angle is <5°; when detecting the neatness of the pillow placement, the feature point recognition rules are the center coordinates of the pillow rectangle, and the quality assessment rules are: the angle between the center line and the edge of the bed is 90°±2°, and the distance between the two centers from the center line of the bed is ±5cm.

[0077] 2) The feature extraction unit is used to identify feature points from a specified image;

[0078] Specifically, identifying feature points from a given image includes the following steps:

[0079] Perform image optimization processing on the specified image, such as distortion correction, reflection removal, multi-scale segmentation and other preprocessing operations, to improve image quality and generate an image to be evaluated, in preparation for subsequent analysis;

[0080] Feature points are identified based on the detection dimensions and standard attributes, including neatness feature points, flatness feature points, and cleanliness feature points;

[0081] Neatness feature points are used to determine whether the pillow and bed sides are aligned and whether the folded corners of the bed sheet meet the standards, such as the center line starting point and the center coordinates of the pillow rectangle.

[0082] Flatness feature points are used to determine whether the wrinkled areas of the bed surface meet the standards;

[0083] Cleanliness feature points are used to determine whether there is damage or foreign objects in the area corresponding to the range type.

[0084] 3) The feature comparison unit is used to calculate the detection result based on the feature points:

[0085] The calculation of test results includes the following steps:

[0086] Select the detection dimension, and extract feature points and their 3D positions from the image to be evaluated based on the detection dimension;

[0087] Standard attributes are extracted based on the detection dimension: If the detection dimension is neatness detection, the standard attributes also include the offset standard value of the bed-making quality detection target, for example, the error value of the angle between the center line and the edge of the bed is ±2°; if the detection dimension is flatness detection, the standard attributes also include the fold density value and fold size value of the bed-making quality detection target; if the detection dimension is cleanliness detection, the standard attributes also include the size value and quantity of foreign objects in the bed-making quality detection target.

[0088] Quality inspection is performed based on standard attributes, feature points, and detection dimensions. Specifically, during neatness inspection, based on feature points and combined with algorithms such as edge extraction and contour fitting, the position of bed sheets, duvet covers, and pillowcases is checked for alignment, the degree of offset and skew is quantified, and the overall neatness of the bed is judged.

[0089] During the flatness detection, based on the lighting rendering model and deep learning network, the wrinkled areas on the surface of the bedding are segmented and extracted, and the wrinkle density, size, and depth are statistically analyzed to evaluate the flatness of the surface. If a suspected wrinkled area is detected, the area is used as a flatness feature point and a focal point to redetermine the image range information. This can be used as a target for obtaining environmental information or as a target for robot smoothing operations.

[0090] During cleanliness inspection, a deep learning-based image semantic segmentation method is used to automatically locate and identify various stains (oil stains, water stains, dirt, etc.) and damages (holes, snags, etc.) on the fabric surface. Hyperspectral analysis is used to determine whether there are foreign objects such as hair or paper scraps remaining on the bedding surface, and the location of foreign objects is marked by a visual detection algorithm. If a suspected foreign object is detected, the area is used as a cleanliness feature point and a focal point to redefine the image range information for further environmental information acquisition or robot grasping, adsorption, and other processing operations.

[0091] 4) Output the inspection results based on feature points and standard attributes. The inspection results include the inspection dimension, focus position, and quality status. For different bedding products, establish quantitative evaluation criteria for key quality factors such as neatness, flatness, degree of soiling, and foreign matter residue, assign scoring weights, and form a quantitative quality index model. The evaluation criterion data is pre-stored as quantitative rules in the control center of the control system. On the other hand, according to the inspection requirements, such as the allowable deviation between business hotels and luxury hotels, the control center also pre-stores dynamically adjusted inspection standards. Specifically, the quality status can be divided into multiple levels such as excellent, good, qualified, and unqualified.

[0092] Step S140: Input image group information into the image recognition model to obtain detection results and correction strategies;

[0093] In practical applications, the complete image group information obtained in step S120 is input into the image recognition model. The image recognition model generates an inspection and detection report and a correction strategy. The correction strategy refers to the further generation of control instructions for the robot, which are used to confirm the detection results and correction quality, as shown in step S141.

[0094] If a suspected foreign object or stain is found during the cleanliness inspection, a correction strategy is generated to confirm the inspection result. At this time, the image range information is redefined and then the action sequence and control instructions are generated based on the image range information. The focus position is the cleanliness feature point, and the sensor takes pictures at multiple angles around the focus position to confirm whether the cleanliness feature point is a foreign object, stain or shadow caused by light.

[0095] If wrinkles are found during the flatness inspection, a correction strategy is generated to correct the quality. At this time, a sequence of actions and control instructions are generated to control the robot's behavior, so that the robot can perform actions such as smoothing out wrinkles or separating contaminated linens.

[0096] In this step, correction strategies can be generated during the inspection process to improve the quality of bed making, controlling the robot to refine the bed making effect during inspection. The inspection ends after the correction strategy is executed a specified number of times, and an inspection report is generated. When generating the inspection report, data visualization technology can be used to transform the inspection results into a quality analysis report containing charts and data. The report marks the type, quantity, and specific location of problematic bedding, clearly identifying the key points requiring rework and rectification.

[0097] The bed-making quality method provided by this invention is integrated into the control system of a hotel bed-making service robot. Through multi-dimensional detection and intelligent analysis, it achieves automated evaluation of bed-making quality, effectively improving the intelligence level of hotel bed-making service robots and providing new ideas for the hotel industry in terms of intelligent upgrading and refined management. Regarding the degree of automation in quality inspection: the quality inspection after robot bed making is fully automated, reducing reliance on manual labor, saving labor costs, and improving the efficiency of hotel room preparation. In particular, the full-item inspection time for a single bed is less than 1 minute, significantly improving inspection speed compared to manual labor. Regarding the accuracy of quality inspection: the introduction of machine vision and intelligent algorithm optimization ensures greater accuracy and consistency in inspection, improving the accuracy of stain identification and the measurement error of positional deviation. Regarding quality management: it improves the readability of evaluation results, providing a quality inspection report that intuitively presents the type, quantity, and location of problematic bedding, facilitating quick location and rectification by the hotel room service management department, and also providing guests with intuitive evidence of bed-making service quality. Regarding robot system control: the generation of control commands during the quality inspection process has also been optimized, reducing unnecessary robot operations. On the other hand, this invention embeds bed-making quality inspection data into all aspects of hotel operations, enabling cross-departmental data sharing and business collaboration. It uses digital means to drive improvements in hotel service quality and optimize operational efficiency. In summary, this invention, based on the cutting-edge needs of robotic applications, specifically develops a bed-making quality inspection system, with targeted designs in perception, analysis, control, evaluation, and optimization, forming a complete technical solution. By ensuring bed-making quality and improving efficiency through digital and intelligent means, this invention has positive significance for promoting the intelligent transformation of the traditional hotel industry and is worthy of widespread application within the industry.

[0098] The above-disclosed embodiments are merely a few specific examples of the present invention. However, the present invention is not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for inspecting the quality of bed making in a hotel bed-making service robot, characterized in that, The bed-making quality inspection method, integrated into the control system of a hotel bed-making service robot, includes the following steps: The quality inspection target for bed making is determined, and the image range information, range type, and inspection dimensions are determined based on the bed making quality inspection target. The range type includes: bed side, sheet fold corner, bed surface, and pillow. The inspection dimensions include: neatness, flatness, and cleanliness. The image range information is determined based on the range type and is the operational characteristic of the robot's sensors when acquiring environmental information. The image range information includes range information identifier, sensor position, focus position, sensor shooting angle, and range type. Control commands for the robot sensors are generated based on the image range information; Obtain the image group information returned by the robot's sensors; the image group information includes: image group information identifier, command identifier, image sequence number, and image file; Load an image recognition model; the image recognition model includes a standard definition unit, a feature extraction unit, and a feature comparison unit; the standard definition unit is used to extract standard attributes from region rules, the standard attributes include the three-dimensional position of the contour edge markers of the bed, feature point recognition rules, and quality assessment rules; the feature extraction unit is used to identify feature points from a specified image; the feature comparison unit is used to calculate the detection result based on the feature points, and supports redefining the image range information during the calculation process; Image group information is input into the image recognition model to obtain detection results and correction strategies; the correction strategies refer to further generating control instructions for controlling the robot, which are used to confirm the detection results and correction quality. The feature extraction unit performs image optimization processing on the specified image to generate an image to be evaluated; it identifies feature points based on detection dimensions and standard attributes, including neatness feature points, flatness feature points, and cleanliness feature points; among them, neatness feature points are used to determine whether the pillow and bed sides are aligned and whether the folded corners of the bed sheet meet the standards; flatness feature points are used to determine whether the wrinkled areas of the bed surface meet the standards; and cleanliness feature points are used to determine whether there is damage or foreign objects in the area corresponding to the range type. The feature comparison unit selects a detection dimension and extracts feature points and their three-dimensional positions from the image to be evaluated based on the detection dimension; it extracts the standard attributes of the quality assessment rules; it performs quality detection based on the standard attributes, feature points, and detection dimension; and it outputs the detection results, which include the detection dimension, focus position, and quality status. Specifically, when the detection dimension is flatness, if a suspected wrinkled area is detected, the area is used as a flatness feature point, and the focus position is used to redefine the image range information. When the detection dimension is cleanliness, if a suspected foreign object is detected, the area is used to generate a cleanliness feature point, and the focus position is used to redefine the image range information.

2. The method for testing the quality of mattress making according to claim 1, characterized in that, The control system of the hotel bed-making service robot controls the behavior of the robot's components through a behavior execution module, a sensor module, and a control module. The components include a mobile base, a robotic arm, and an end effector; The behavior execution module includes an end effector unit, a robotic arm control unit, and a motion control unit, which are deployed on the constituent components to respond to control commands and control the actions of each constituent component. The sensor module includes a vision system unit consisting of an RGB camera and a structured light camera, which is placed on the robot's components and can collect environmental information at a specified location, angle, and pattern. The control module includes an operation specification library, a task planning engine, and a mode switching unit. It is deployed on the control server and is used to plan the specific operation steps for bed making based on environmental information, generate functional instructions, and send control instructions to the behavior execution module. The control command includes: command identifier, execution module ID, operation component ID, target point location, operation type, and control status; When the operation type is shooting, the control status includes camera position, camera orientation, focus position, and shooting parameters.

3. The method for testing the quality of mattress making according to claim 1, characterized in that, When determining the image range information and range type, the target for bed quality detection is decomposed into positions, and a focal position corresponding to the range type is generated according to the region rules; the focal position is determined by three-dimensional coordinates based on the mattress coordinate system; one focal position corresponds to multiple sensor shooting angles.

4. The method for testing the quality of mattress making according to claim 1, characterized in that, Before generating control commands for the robot's sensors, the process of generating control commands is optimized. The optimization goal is to reduce repetitive operations of the robot. The optimization method is to merge the contents of multiple image range information and store the correspondence between control commands and image range information. During optimization, it is calculated whether there is an overlap between the focal positions corresponding to all image range information. If there is an overlap and the sensor shooting angle is the same, the control commands of the image range information are merged into one control command. If there is no overlap, under the premise of ensuring the clarity of each focal position, the degree of concentration of the focal positions is judged, the core concentration point of multiple focal positions with close distance is found, the core concentration point is taken as the optimal focal position, and the sensor shooting angle is added to generate a control command. Other focal positions with close distance do not need to generate control commands.

5. The method for testing the quality of mattress making according to claim 4, characterized in that, After obtaining the image group information returned by the robot sensor, the complete image group information that completely corresponds to the image range information is restored according to the correspondence between the control command and the image range information.

6. The method for testing the quality of mattress making according to claim 1, characterized in that, If the detection dimension is uniformity, the standard attributes also include the offset standard value of the bed-making quality detection target; if the detection dimension is flatness, the standard attributes also include the fold density value and fold size value of the bed-making quality detection target; if the detection dimension is cleanliness, the standard attributes also include the size and quantity of foreign objects in the bed-making quality detection target.

7. The method for testing the quality of mattress making according to claim 1, characterized in that, When confirming the detection result, the correction strategy is to redetermine the image range information and then generate an action sequence and control instructions based on the image range information; When correcting the quality, the correction strategy is to generate a sequence of actions and control instructions to control the robot's behavior, so as to control the robot to perform actions such as smoothing out wrinkles or separating soiled linens.

8. The method for testing the quality of mattress making according to claim 1, characterized in that, The quality inspection ends after the correction strategy is executed a specified number of times, and an inspection report is generated. The inspection report marks the type, quantity, and specific location of the problematic bedding, and clarifies the key points that need to be reworked and rectified.

Images